Thermal image enhancement using kurtosis based clipping histogram method

Enhancement is necessary to improve the visibility of the thermal or infrared images useful in the surveillance applications. As the thermal images will suffer with poor visibility, low contrast and low resolution it is difficult to enhance. In this paper, two new approaches are introduced for increasing the visibility of the low illumination thermal infrared images. The first method is based on the Histogram matching technique. The histogram of the low illumination input image is modified in accordance with the statistical parameters of reference image. Second approach is based on clipping the histogram using kurtosis and then bisecting the histogram using the median. For performance evaluation few images from OSU thermal dataset are used and compared with existing methods. The proposed technique performance is evaluated using various quality metrics such as PSNR, SSIM, Entropy and Absolute Mean Brightness Error (AMBE). The results have shown the improvement in the PSNR, SSIM and AMBE when compared with the mean based clipping. The performance of the proposed methods gave improved quantitative parameters in comparison with the state of the art methods.